The REG X8 memory is registered ECC memory, the other is not.
The X8 is in reference to the density; Paul writes a great piece which explains in depth the density and thus the X8 and X16 references.
At the 512Megabit level, some chip
types might be 16Mx16, 32Mx8, 64Mx4.
The "x16" and "x8" types are defined
as acceptable by JEDEC, for making
unbuffered modules. With the "x16"
chips, four of those chips on one side
of the module, makes a bank.
With the "x8" type, it takes eight of
those chips on the side of a module,
to make a bank. Basically, it takes as
many chips as is necessary to make a
64 bit grouping (as the bus interface
on a DIMM is 64 bits wide).
X8 would be a low density chip, so make sure your motherboard supports it. The X8 is in reference to the memory configuration on the chip itself.
In terms of better; that is subjective. One is geared towards scaling and addresses technical limitations of scaling by way of ECC.
You should see if your server currently has ECC or non-ECC memory in it. If it has low density ECC memory, go with that, otherwise go with the non-ECC memory.
I understand the "registered" part- is it better than non-registered? what about the "x8"? – sman591 – 2011-05-29T03:53:06.623
@sman591 Better is subjective; ECC is slower than non-ECC memory; answer updated. – Aaron McIver – 2011-05-29T21:26:41.327
@sman591 On the other hand, ECC provides better data integrity than standard memory. – AndrejaKo – 2011-05-29T23:01:01.777